[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-bugs-dist
Subject:    [Bug 52180] kswapd starts thrashing when mozilla konquorer and
From:       Oliver Tennert <tennert () science-computing ! de>
Date:       2003-04-12 10:38:01
[Download RAW message or body]

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.
     
http://bugs.kde.org/show_bug.cgi?id=52180     




------- Additional Comments From tennert@science-computing.de  2003-04-12 12:37 -------
PARTIAL RELIEF:

OK, as nobody else seems to be afflicted (or interested?) by this memory leak 
problem, I have done some further research:

1.) There is __definitely__ no konsole problem, but a kdeinit problem. The 
thrashing occurs with applications like khelpcenter oder konqueror as well.

2.) This is a typical valgrind output (starting konqueror):

...
==2149== Syscall param modify_ldt(ptr)(func=1 or 0x11) contains uninitialised or 
unaddressable byte(s)
==2149==    at 0x41473948: (within /usr/lib/GL/libGL.so.1.0.4349.nv_glx)
==2149==    Address 0xBFFFF3D0 is on thread 1's stack
==2149==
==2149== Syscall param writev(vector[...]) contains uninitialised or 
unaddressable byte(s)
==2149==    at 0x4120A27D: my_do_syscall3 (in /usr/lib/valgrind/libpthread.so)
==2149==    by 0x412094E2: vgIntercept_writev (in /usr/lib/valgrind/libpthread.
so)
==2149==    by 0x401707E4: __writev (in /usr/lib/valgrind/valgrind.so)
==2149==    by 0x4116C74F: _X11TransSocketWritev (in /usr/X11R6/lib/libX11.so.6.
2)
==2149==    Address 0x44109493 is 207 bytes inside a block of size 2048 alloc'd
==2149==    at 0x4015E999: calloc (in /usr/lib/valgrind/valgrind.so)
==2149==    by 0x4113E278: XOpenDisplay (in /usr/X11R6/lib/libX11.so.6.2)
==2149==    by 0x40C0D114: qt_init_internal(int*, char**, _XDisplay*, unsigned 
long, unsigned long) (in /usr/lib/qt3/lib/libqt-mt.so.
3.1.2)
==2149==    by 0x40C0DE8D: qt_init(int*, char**, QApplication::Type) (in 
/usr/lib/qt3/lib/libqt-mt.so.3.1.2)
==2149== valgrind's libpthread.so: UNIMPLEMENTED FUNCTION: 
__libc_current_sigrtmin
==2149==
valgrind's libpthread.so: Please report this bug to me at: jseward@acm.org


==2149== discard syms in /usr/lib/gconv/ISO8859-1.so due to munmap()
==2149== discard syms in /lib/libnss_files.so.2 due to munmap()
==2149==
==2149== ERROR SUMMARY: 5 errors from 2 contexts (suppressed: 237 from 6)
==2149== malloc/free: in use at exit: 2119967 bytes in 38089 blocks.
==2149== malloc/free: 267983 allocs, 229894 frees, 8538155 bytes allocated.
==2149== For a detailed leak analysis,  rerun with: --leak-check=yes
==2149== For counts of detected errors, rerun with: -v

A similar output is obtainable by starting konsole, kinternet oder anything 
else.

3.) Now the interesting thing: there certainly __is__ a memory leak somewhere, 
but for the thrashing to be occurring, a buggy glibc is needed, too!! I had RPM 
package "glibc-2.2.5-165.i686.rpm" from the updates directory of "ftp.suse.com", 
which seems broken. Downgrading to another version (an "i586" package) leads to 
a stable system, even with swap activated.

4.) Now there is no thrashing anymore, but it often occurs that KDE applications 
are not able to connect to DCOM server, WHENEVER SWAP SPAVE IS ACTIVATED. 
Starting programs like kmail or anything else leads to a segmentation fault. 
Deactivating swap space has you get rid of any problems!

5.) CONCLUSION: kdeinit still has problems when swap space is activated (an 
issue which might be connected to the VM subsystem of the kernel and the 
different behaviour with or w/o swap), but taking a different glibc at least 
gets you a stable system without thrashing!!

******************************
[prev in list] [next in list] [prev in thread] [next in thread]